Envelopes and evolutes

Abstract

The study of evolutes of plane curves goes back at least to Huygens, and was continued and extended to space curves by Monge, Darboux, and others. Salmon studied projective curves and surfaces and their evolutes and gave many enumerative formulas for their degrees and number of singularities. We define envelopes of families of linear spaces in projective space. In order to define evolutes we impose a notion of perpendicularity, which allows us to consider the normal spaces to a projective variety. The evolute of a projective hypersurface is the envelope of the family of normal lines. For a variety of dimension $r$ in $n$-space, the evolute is defined as the $(n-r)$th iterated cuspidal locus of the map from the total space of the normal spaces to projective space. Thus the envelope can be interpreted as a $(n-r)$th order Thom-Boardman singularity. Further higher order Thom-Boardman singularities correspond, for a curve in the plane or in 3-space, to classical objects like the vertices of the curve; for a surface in 3-space, they give the cuspidal curve - and its cusps - on the evolute. Using known formulas for Thom polynomials we are able to verify and generalize many of Salmon's formulas and find new ones.